Research On Runoff Forecasts With Different Lead Times And Optimization Operation For Hydropower Station

At present, our country are taking vigorous actions to develop new energy and renewable energy construction, as a clean energy, the hydropower has a unique advantage. There are usually two ways to increase energy. The first one is to construct new hydropower stations, and the second one is to develop effective operation systems for the existed hydropower stations using the runoff forecasts to enhance the utilization ratio of water resources, without increasing the project investment. Due to the environmental issue, it becomes more difficult to construct new hydropower stations, therefore, the second method has been an important way to increase energy. Although the medium-and short-term runoff forecasts have shown an important value in the hydropower station operation, a further research on the use of the long-term runoff forecast is still needed to improve the efficiency of power generation. Therefore, with Huanren hydropower station as a case study, the runoff forecast using the ensemble precipitation forecasts from ECMWF, the uncertainty in the medium-term runoff forecast, the practicability of decomposition-combination models and the long-term forecast of the runoff during the remaining flood season of every period are first studied in this dissertation, and then a hydropower station optimization operation model coupling the runoff forecasts with different lead times is developed. The main content and research results are as follows:(1) The precipitation products can be used as the alternative input of rain-runoff forecast model in the basins where the precipitation is lack, and the precipitation products can be used to extend the lead time of runoff forecasts, however, the suitability evaluation and availability analysis are needed before their applications. The substitutability and suitability of three precipitation products (GLDAS, TMPA, ERA-Interim) are first evaluated, and then the optimum selection model are developed for the selection of the precipitation products. After that, the availability analysis of the ensemble precipitation forecasts from ECMWF are carried out. Results reveal that, TMPA is more suitable for hydrological runoff modelling at the basin scale, ERA-Interim is less suitable, and GLDAS cannot be used as the precipitation alternative due to the underestimate of the observed precipitation, which will bring risk to the flood control, the mean precipitation of the ensemble members can be used for runoff forecast.(2) To quantify and decompose uncertainties in the medium runoff forecast, a decomposition scheme is proposed using the analysis of variance (ANOVA) method. First, three data-driven models (i.e., ANN, SVM, ANFIS) and six variables (i.e., inflow (Q), precipitation (P), relative humidity (H), minimum temperature (Tmin), maximum temperature (Tmax) and precipitation forecast (F)) are used to produce an ensemble of 10-day inflow forecast for Huanren hydropower station in China, and ANOVA method is employed to decompose the uncertainty. The ensemble forecast results show that when the three variables, i.e., Q, P and F, are used only, the predictive accuracy of the data-driven models is very high, and the loss of accuracy without the other three, i.e., H, Tmin and Tmax, is very small. The decomposition results indicate that the input set is the dominant source of uncertainty, the contribution of the data-driven model is limited and has a strong seasonal variation:larger in winter and summer, smaller in spring and autumn. Most importantly, the interactive contribution of the input set and the data-driven model to the total predictive uncertainty is very high and is more significant than the individual contribution from the model itself.(3) To evaluate the practicability of the decomposition-combination models, a validation scheme including hindcast experiment and forecast experiment is proposed in this study. First, six hybrid models are developed in this study, including WA-ANN, WA-ARMA, EMD-ANN, EMD-ARMA, SSA-ANN and SSA-ARMA, with the combination of three preprocessing techniques, such as wavelet analysis (WA), empirical mode decomposition (EMD) and singular spectrum analysis (SSA), and two modeling methods (i.e. ANN and ARMA). Then the hindcast and forecast experiments are designed for investigating the performance of the hybrid models and the impact of the extension methods on the performance of the WA-based and EMD-based models. Results indicate that the six hybrid models perform worse in the forecast experiment compared with the original ANN and ARMA models, leading to the conclusion that the hybrid models are not suitable for runoff forecasting in this study, while the hybrid models in the hindcast experiment perform better than the original models due the advance use of future information, and the performances of WA-based and EMD-based models vary largely across different extension methods. New extension methods and modified preprocessing techniques can improve the prediction performance of these hybrid models in forecast experiments.(4) The real-time long-term runoff forecast models are developed to forecast the runoff during the remaining flood season of every period. First, the flood season (from May to October) is divided into 5-day periods, and then the long-term forecast model for the runoff during the remaining flood season of every period is developed using ANN models and atmospheric circulation factors. Results indicate that both the qualified rate and grade estimation of 80% periods in flood seasons are more than 80%, revealing the forecast models show a satisfactory performance. And the developed models are effective method for forecasting the runoff in flood season, because they not only can count for the impact of the atmospheric circulation factors on the runoff during the remaining flood season, but also is friendly to use. The forecast results provide the long-term runoff forecasts required by the reservoir operation model.(5) Based on the achievements in the application of medium-term runoff forecasts in the operation of the typical hydropower station, the employment of the long-term runoff forecast is studied, and a hydropower station operation model coupling runoff forecasts with different lead times (LMS-BSDP) is developed. First, the availability of runoff forecasts with different lead times is analyzed, which including the first 5-day forecast and the second 5-day forecast from the medium-term runoff forecasts with 10 days, as well as the long-term forecast for the runoff during the remaining flood season of the second 5 days, then, the three runoff forecasts are coupled by the Bayesian theory and the LMS-BSDP model is developed, finally, the new developed model is viladated. The simulation results demonstrate that the performance of LMS-BSDP is superior to that of TS-BSDP and other models.